|
|||||||||
Allow me to consider myself in virtual computer reality to be a member of a grandmaster team in preparation of a thematic demonstration contest, “Man Against Machine” in real time. One million dollars/euros are at stake. Having studied the high risk games of Topalov and Cheparinov for quite some time now, I think that these eminent Bulgarian contemporaries have the appropriate sense of adventure. What would it take? First of all, do NOT waste time and energy playing Black in the 1.d4 domain. Draw the d4-games with Black, head for the bar and call it a day. Respond with 1...c5 to white’s 1.e4. The Open Sicilian ALWAYS, but NOT the g6-systems. Use the tempi for g6/Bg7/0-0 for a swift advance on the queenside if the machine chooses an asymmetric 0-0-0/0-0 position. Probably the only chance for Black in computer chess. And NO Svesnikov e5-lines. Sicilian lines have a tendency to evolve into positions in which Black can operate with the necessary lightheartedness. With White of course we play 1.e4 to provoke the open Sicilian again. In case the machine responds to 1.e4 with 1...e5, DRAW the game with a Ruy Lopez or a Petroff, 1.e4 e5 2.Nf3 Nf6. This might offend spectators and sponsors, but my advice is to adhere strictly to these principles. Computers defend the typical Russian variations (Archangelsk, Kalashnikov,Karpov-Zaitsev, Chigorin, Keres, etc.,etc) with a moderate White advantage effectively. The Marshall Attack is 100 years old and it is analyzed in a multitude of lines up to move 30. It is the best proof we have at the moment that the noble game is a draw when played without calculative errors. Use Kasparov’s anti-Marshall line, if need be. We can venture Italian, Two Knights or the Vienna opening(!) in 1.e4 e5 games, but NOT before we have a couple of Sicilian Black & White victories under our belt. That leaves us with the French, Aljechin and Caro-Kann. etc. I have played Fritz-10, I estimate several thousands times. INVARIABLY it responds to 1.e4 Nf6 (Aljechin) 2.d4 (NOT 2.e5) with 2...d6 (never 2...Ne4) 3.Nf3 c5 and we transform to a d6-Sicilian. French? On 1.e4 e6 not 2.d4 but 2.Nf3 followed by 2...c5 3. d4 cd4 4.Nd4 and we switch to an e6-Sicilian. We can talk about 1.e4 c6 some other time. In other words: we focus on a limited set of sharp Sicilian variations with a narrow draw margin and explore them in as much detail and as far as possible. When playing machines, play like a machine. They have no mercy and neither should we. Spectacular forced combinatory wins within 20-30 moves are a rarity in computer chess, and in tournaments they are due to weaker human continuations. Computers “play” with perfect calculation. However, this is a weakness too. It makes them very predictable (reproducible is the scientific term) and we do not have to refute unexpected human sidelines. In general one ends up in a better endgame that requires precise “technical “ execution. A professional grandmaster should be able to memorize specialized preparations. Put the first say, 20 moves, on the board in high tempo. The computer has the maddening habit to respond INSTANTANEOUSLY on all human moves coming up. We do not want to waste the precious commodity called time in the opening ceremonies. ******* Dwelling in the Sicilian jungle with its eternal vendetta scenario, let me first explain why we should not play g6-systems with Black. For White:
The configurations with 0-0-0 for White and 0-0 for Black install long term positional asymmetry without immediate material sacrifice. Besides: storming each other's castles in high tempo is GREAT FUN. Two continuations A. 8...0-0 and B. 8...Qb6: A. 8...0–0 (in book and without book too).
Now for... B. 8...Qb6! (A nasty human book move).
Funny position. FIVE black pawns on the 6th row. White is slightly better but is it technically possible to convert it to a win, at the board, in limited time? We have to work on that one. In the event of 1.e4 Nf6 2.d4 d6 3.Nf3 c5 4.Nc3 cxd4 5.Nxd4 g6 6.Bc4 Nc6 7.f3 Bg7 8.Be3 0–0 9.Qd2 Bd7 10.0–0–0 Rb8 (book) white should go 11.Bb3(N) because black threatens to play b5 and sacrificial lines to open fire on the white castle: 11.- Na5 12.Bh6 Bxh6 13.Qxh6 Rc8 14.Kb1 Nxb3 15.cxb3 Qa5 16.g4 a6 17.h4 Kh8 18.Nd5 Nxd5 19.exd5 Qd8 20.Rde1 Rg8 21.Re4 e5 22.dxe6 fxe6 23.Nxe6 Qf6 24.Rhe1 d5 25.Re5 b6 26.Ng5 Qg7 27.Qxg7+ Rxg7 28.Rxd5 and WIN. After neutralization of the g7-bishop, white rules the black squares on the front lawn of the fianchetto castle. We will discuss playing Sicilian with Black in an upcoming session. I have investigated a multitude of Sicilian 0-0-0/0-0-positions. Black has to use ALL his material, open files via sacrifice of pawn(s) or construe the exchange of the white dark-squared bishop for a rook or trade a rook for a white knight on c3 (Shirov-Anand, Linares 2008, played without 0-0 for Back!). In the Dragons the configuration with g6/Bg7/0-0 decelerates the Black advance and provokes the typical Velimirovic attacks on the weakened kingside. In the non-g6 lines, the computer will get busy defending its queenside castle and launch its own attack too late. Or not at all. Now for White: Sicilian Najdorf 5...a6 with early 6.f4. Regular or via a transformed Aljechin Defense:
In the majority of Bg5-lines, White ends up exchanging the bishop at f6. In computer chess, bishops are definitely more effective then knights, in particular a bishop pair. Very often a bishop outweighs a rook in an open position. Therefore I opt for the rather restrained Scheveningen Najdorf with early 6.f4 with Be2 and Be3. In the book is now 13.f5 followed by 13...b4. I experimented extensively with the sacrifice 13.Nf5!? in Topalov-style. But it can be defended: 13.Nf5!? exf5 14.Nd5 Qd8 15.exf5 Re8 16.Bf3 Bb7 17.Qg2 Bf8 18.f6 Kh8 19.Rhe1 Qa5 20.Ne7 Bxe7 21.Bxc6 Bxc6 22.Qxc6 Nxf6 23.gxf6 Bxf6 24.Bd2 Qd8 25.Ba5 Qb8 26.Rxd6 Rxe1+ 27.Bxe1 Be7 28.Rd7 Qe8 29.Qe4 Bc5 30.Qxe8+ Rxe8 to a draw. Exemplary of the conclusion that rather brute force light piece sacrifices in the first, say, 15 moves are not effective against computers. So we have:
Black can not allow the g-file to be opened up. After 14.f6 the c3-knight is accepted:
This one starts at a healthy advantage, a bishop pair up against a rook and a pawn. Ten moves to go until the time control. White is good here. From the stem diagram:
Presumably all the lines after the book move 13...b4 are White wins. That makes me suspicious. Will Fritz-10 do that one with the book switched off? NO! In my upcoming monograph How To Fool Fritz II, I discovered DOZENS of examples of wins over computer programs due to erratic human instruction. When the book is switched off at a critical point, one will note that the machines invariably improves on the human book move(s). So instead of 13...b4 in...
...the machine defends with 13...Nxd4 and now the sacrifice on c3 after 13.Bxd4 b4 is ineffective. In many lines a sturdy Svesnikov-like e5 by Black obstructs the White plans as well. We have the rather prosaic:
And a firm win at move 40. It is unique move sequence after 25...Bd4 reproduced time and time again. I found the lines with 14.Qd4 or a preceding 13.Nc6 to avoid the exchange on d4 to converge on queen endings with too many variables. We close out with an exchange of a white rook for a black bishop. First, we fence off the b4-advance with 13.a3.
Lore has it that spectators showered the board with gold coins after Frank J. Marshall won via a classic queen sacrifice in Breslau 1912 (www.howtofoolfritz.com Tim Krabbe's Collection of 100 Best Chess Moves Ever). This last one certainly qualifies! But as usual, once bitten twice shy and the machine construes defenses with Nc5 or Rb8 leaving the bishop on the c8-f5 diagonal and refusing and the rook on h1 in subsequent runs. In conclusion: we have two options in a real time Man Against Machine match: (1) we invent novelties in computer improved book lines or (2) we beat the anonymous computer crew book compilers. First come, first served. Take the money and run! Albert
Alberts Albert Alberts' Explorations in Man-Assisted Machine Chess
|
Advertisement
The
Advertise to Single insert:
|
||||||||
|
|||||||||
“The
problem with chess computers is TIME. Without time limit, humans will
still win.” A quote from Veselin Topalov transcribed in
